Older adults with dementia are at increased risk of hospitalization when compared to adults without dementia of similar age and medical comorbidity. The increased risk of hospitalization extends to potentially preventable hospitalization (PPH) for conditions such as a urinary tract infection or asthma exacerbation, suggesting difficulty in outpatient management of patients with dementia. Neuropsychiatric symptoms (NPS) of dementia such as agitation or delusions likely account for a significant amount of this risk, given their prevalence and potential to cause caregiver distress. While there are effective interventions for patients and caregivers to reduce NPS, the profile of patients that could benefit the most from intervention, therefore reducing their hospitalization risk, is unknown. Through the coordinated program of mentorship, didactics, and research that I propose, I will develop the advanced skills to derive and apply administrative, claims, and clinical encounter data to prospectively identify those patients with dementia at highest risk for hospitalization. Development of this patient-level risk phenotype means that future interventions to reduce hospitalization can then be prospectively matched to the patients most likely to benefit, a development of critical public health importance given both financial and geriatric work force constraints. Over the next four years, my short-term training goals include: (1) address gaps in my formal research training, specifically: (a) to conduct observational analyses using large-scale claims and administrative data; (b) to derive clinical data from the electronic health record using natural language processing; and (c) to apply advanced methods of data analysis for risk prediction; (2) train in presentations, manuscript writing, and grantsmanship that culminate with a R01 proposal; (3) establish further connections with potential collaborators in the University of Michigan (UM) Pepper Center and broader community of aging researchers, national geriatrics and geriatric psychiatry communities, and the Beeson Scholar community; and (4) engage in leadership development with an emphasis on skills to lead a research team, mentor junior investigators, and communicate findings in research and clinical care settings. These short-term goals will be paired with research aims that focus on elaborating the PPH risk profile for patients with dementia. Such research objectives can only be achieved when: (1) full clinical characteristics are available for the at-risk (i.e., non-hospitalized) population, includig (2) NPS data, which are rarely captured in standard administrative claims data. These criteria are uniquely met in the Veterans Affairs healthcare system, which has one of the nation's most advanced electronic health records (EHR). Using a national dementia case repository (N=269,565) from which I will draw matched cases (patients with dementia + PPH) and controls (non-hospitalized patients with dementia). Aim 1 will use claims and administrative data to explore patient, treatment, and facility risk factors associated with PPH. Aim 2 will use natural language processing to derive NPS from EHR clinical encounter notes and then characterize the association of NPS with PPH. Using the risk phenotype described in Aims 1 and 2, Aim 3 will develop logistic risk-prediction models to prospectively identify patients with dementia at highest risk for PPH. In subsequent grant proposals I will validate this risk- prediction model in other healthcare systems and prospectively pair the assessment tool with an evidence- based dementia intervention to reduce hospitalization. My long-term career goals are to: (1) establish myself as independent investigator and national leader in geriatric mental health services research; (2) develop a programmatic line of funded health services research that develops risk-stratification models for late-life mental health and cognitive disorders; (3) translate knowledge from these research endeavors to improve the targeting and impact of future interventions research and health system delivery strategies; and (4) contribute broadly to the care of older adults by training and mentoring future clinical researchers in late-life mental health disorders. I am an Assistant Professor and geriatric psychiatrist at the University of Michigan, where I am also currently completing a MSc in Health and Healthcare Research, which provides an excellent background in health services research for clinicians. With this combination of clinical expertise and foundational training in health services research, I am uniquely qualified to undertake the advanced training activities outlined in this proposal, while UM affords the ideal environment in which to pursue this work. My primary mentor (Helen Kales, MD) and co-mentor (Frederic Blow, PhD) are national leaders in geriatric mental health who have used observational data to answer questions of national significance. My Advisory Panel includes Constantine Lyketsos, MD, MHS, an internationally-recognized expert in NPS and dementia care, and Kenneth Langa, MD, PhD, an internist, former Beeson Scholar, and renowned expert in using survey and secondary data to inform our understanding of dementia. Consultants include David Hanauer, MD, MS, an expert in bioinformatics and natural language processing, and Rodney Hayward, MD, a leader in risk assessment and intervention- targeting. My advisory team paired with resources of Michigan's Pepper Center, CTSA, and multi-disciplinary Institute for Healthcare Policy and Innovation make this the ideal environment in which to complete the proposed training activities.